|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use PiedWeb\UrlHarvester\Harvest; |
|
6
|
|
|
use PiedWeb\UrlHarvester\Indexable; |
|
7
|
|
|
use PiedWeb\UrlHarvester\Link; |
|
8
|
|
|
|
|
9
|
|
|
class Crawler |
|
10
|
|
|
{ |
|
11
|
|
|
const FOLLOW = 1; |
|
12
|
|
|
const NOFOLLOW = 2; |
|
13
|
|
|
|
|
14
|
|
|
protected $recorder; |
|
15
|
|
|
|
|
16
|
|
|
protected $robotsTxt; |
|
17
|
|
|
protected $request; |
|
18
|
|
|
|
|
19
|
|
|
protected $currentClick = 0; |
|
20
|
|
|
|
|
21
|
|
|
protected $counter = 0; |
|
22
|
|
|
|
|
23
|
|
|
protected $urls = []; |
|
24
|
|
|
|
|
25
|
|
|
/** |
|
26
|
|
|
* @var CrawlerConfig |
|
27
|
|
|
*/ |
|
28
|
|
|
protected $config; |
|
29
|
|
|
|
|
30
|
|
|
public function __construct( |
|
31
|
|
|
string $startUrl, |
|
32
|
|
|
string $ignore, |
|
33
|
|
|
int $limit, |
|
34
|
|
|
string $userAgent, |
|
35
|
|
|
int $cacheMethod = Recorder::CACHE_ID, |
|
36
|
|
|
int $wait = 100000, // microSeconds ! |
|
37
|
|
|
?string $dataDirectory = null |
|
38
|
|
|
) { |
|
39
|
|
|
$this->config = new CrawlerConfig($startUrl, $ignore, $limit, $userAgent, $cacheMethod, $wait, $dataDirectory); |
|
40
|
|
|
|
|
41
|
|
|
$this->urls[$this->config->getStartUrl()] = null; |
|
42
|
|
|
|
|
43
|
|
|
$this->recorder = new Recorder($this->config->getDataFolder(), $this->config->getCacheMethod()); |
|
44
|
|
|
|
|
45
|
|
|
$this->config->recordConfig(); |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
public function getConfig(): CrawlerConfig |
|
49
|
|
|
{ |
|
50
|
|
|
return $this->config; |
|
51
|
|
|
} |
|
52
|
|
|
|
|
53
|
|
|
public function crawl(bool $debug = false) |
|
54
|
|
|
{ |
|
55
|
|
|
$nothingUpdated = true; |
|
56
|
|
|
|
|
57
|
|
|
if ($debug) { |
|
58
|
|
|
echo PHP_EOL.PHP_EOL.'// -----'.PHP_EOL.'// '.$this->counter.' crawled / ' |
|
59
|
|
|
.count($this->urls).' found '.PHP_EOL.'// -----'.PHP_EOL; |
|
60
|
|
|
} |
|
61
|
9 |
|
|
|
62
|
|
|
foreach ($this->urls as $urlToParse => $url) { |
|
63
|
|
|
if (null !== $url && (false === $url->can_be_crawled || true === $url->can_be_crawled)) { // déjà crawlé |
|
64
|
|
|
continue; |
|
65
|
|
|
} elseif ($this->currentClick > $this->config->getLimit()) { |
|
66
|
|
|
continue; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
if ($debug) { |
|
70
|
9 |
|
echo $this->counter.'/'.count($this->urls).' '.$this->config->getBase().$urlToParse.PHP_EOL; |
|
71
|
9 |
|
} |
|
72
|
9 |
|
|
|
73
|
9 |
|
$nothingUpdated = false; |
|
74
|
9 |
|
++$this->counter; |
|
75
|
9 |
|
|
|
76
|
9 |
|
if (null === $this->urls[$urlToParse]) { |
|
77
|
|
|
$url = $this->urls[$urlToParse] = new Url($this->config->getBase().$urlToParse, $this->currentClick); |
|
78
|
9 |
|
} |
|
79
|
|
|
|
|
80
|
9 |
|
$harvest = false === $this->canBeCrawled($url) ? null : $this->harvest($url); |
|
81
|
|
|
$this->urls[$urlToParse]->setDiscovered(count($this->urls)); |
|
82
|
9 |
|
|
|
83
|
9 |
|
$this->cacheRobotsTxt($harvest); |
|
84
|
9 |
|
|
|
85
|
9 |
|
$this->cacheRequest($harvest); |
|
86
|
9 |
|
|
|
87
|
9 |
|
usleep($this->config->getWait()); |
|
88
|
9 |
|
|
|
89
|
9 |
|
if ($this->counter / 500 == round($this->counter / 500)) { |
|
90
|
|
|
echo $debug ? ' --- auto-save'.PHP_EOL : ''; |
|
91
|
9 |
|
$this->recorder->record($this->urls); |
|
92
|
|
|
} |
|
93
|
|
|
} |
|
94
|
|
|
|
|
95
|
|
|
++$this->currentClick; |
|
96
|
9 |
|
|
|
97
|
|
|
// Record after each Level: |
|
98
|
9 |
|
$this->recorder->record($this->urls); |
|
99
|
9 |
|
|
|
100
|
|
|
$record = $nothingUpdated || $this->currentClick >= $this->config->getLimit(); |
|
101
|
6 |
|
|
|
102
|
|
|
return $record ? null : $this->crawl($debug); |
|
103
|
6 |
|
} |
|
104
|
|
|
|
|
105
|
|
|
protected function cacheRobotsTxt($harvest) |
|
106
|
9 |
|
{ |
|
107
|
|
|
if (null === $this->robotsTxt && $harvest instanceof Harvest) { |
|
108
|
9 |
|
$this->robotsTxt = $harvest->getRobotsTxt(); |
|
109
|
|
|
} |
|
110
|
|
|
|
|
111
|
|
|
return $this; |
|
112
|
9 |
|
} |
|
113
|
9 |
|
|
|
114
|
|
|
protected function cacheRequest($harvest) |
|
115
|
9 |
|
{ |
|
116
|
|
|
if ($harvest instanceof Harvest && null !== $harvest->getResponse()->getRequest()) { |
|
117
|
|
|
$this->request = $harvest->getResponse()->getRequest(); |
|
118
|
9 |
|
} |
|
119
|
|
|
|
|
120
|
9 |
|
return $this; |
|
121
|
|
|
} |
|
122
|
|
|
|
|
123
|
9 |
|
protected function loadRobotsTxt(Harvest $harvest) |
|
124
|
|
|
{ |
|
125
|
9 |
|
if (null !== $this->robotsTxt) { |
|
126
|
|
|
$harvest->setRobotsTxt($this->robotsTxt); |
|
127
|
9 |
|
} |
|
128
|
9 |
|
|
|
129
|
9 |
|
return $this; |
|
130
|
|
|
} |
|
131
|
|
|
|
|
132
|
9 |
|
protected function getHarvester(Url $url) |
|
133
|
9 |
|
{ |
|
134
|
3 |
|
return Harvest::fromUrl( |
|
135
|
9 |
|
$this->config->getBase().$url->uri, |
|
136
|
2 |
|
$this->config->getUserAgent(), |
|
137
|
|
|
'en,en-US;q=0.5', |
|
138
|
|
|
$this->request |
|
139
|
9 |
|
); |
|
140
|
9 |
|
} |
|
141
|
|
|
|
|
142
|
|
|
protected function canBeCrawled(Url $url) |
|
143
|
9 |
|
{ |
|
144
|
9 |
|
if (null === $url->can_be_crawled) { |
|
145
|
|
|
$url->can_be_crawled = $this->config->getVirtualRobots() |
|
146
|
9 |
|
->allows($this->config->getBase().$url->uri, $this->config->getUserAgent()); |
|
147
|
9 |
|
} |
|
148
|
|
|
|
|
149
|
9 |
|
return $url->can_be_crawled; |
|
150
|
|
|
} |
|
151
|
9 |
|
|
|
152
|
|
|
protected function harvest(Url $url): ?Harvest |
|
153
|
9 |
|
{ |
|
154
|
|
|
$harvest = $this->getHarvester($url); |
|
155
|
9 |
|
|
|
156
|
|
|
if (!$harvest instanceof Harvest) { |
|
157
|
6 |
|
$url->indexable = Indexable::NOT_INDEXABLE_NETWORK_ERROR; |
|
158
|
|
|
|
|
159
|
|
|
return null; |
|
160
|
|
|
} |
|
161
|
9 |
|
|
|
162
|
|
|
$this->loadRobotsTxt($harvest); |
|
163
|
|
|
|
|
164
|
9 |
|
$url->indexable = $harvest->isIndexable(); // slow ~30% |
|
165
|
|
|
|
|
166
|
9 |
|
if (Indexable::NOT_INDEXABLE_3XX === $url->indexable) { |
|
167
|
|
|
$redir = $harvest->getRedirection(); |
|
168
|
9 |
|
if (false !== $redir) { |
|
169
|
|
|
$links = Harvest::LINK_INTERNAL === $harvest->getType($redir) ? [new Link($redir)] : []; |
|
170
|
|
|
} |
|
171
|
9 |
|
} else { |
|
172
|
|
|
$this->recorder->cache($harvest, $url); |
|
173
|
9 |
|
|
|
174
|
6 |
|
$mimeType = $harvest->getResponse()->getMimeType(); |
|
175
|
|
|
$url->mime_type = 'text/html' == $mimeType ? 1 : $mimeType; |
|
176
|
|
|
|
|
177
|
9 |
|
$this->recorder->recordOutboundLink($url, $harvest->getLinks()); // ~10% |
|
178
|
|
|
$url->links = count($harvest->getLinks()); |
|
179
|
|
|
$url->links_duplicate = $harvest->getNbrDuplicateLinks(); |
|
180
|
9 |
|
$url->links_internal = count($harvest->getLinks(Harvest::LINK_INTERNAL)); |
|
181
|
|
|
$url->links_self = count($harvest->getLinks(Harvest::LINK_SELF)); |
|
182
|
9 |
|
$url->links_sub = count($harvest->getLinks(Harvest::LINK_SUB)); |
|
183
|
6 |
|
$url->links_external = count($harvest->getLinks(Harvest::LINK_EXTERNAL)); |
|
184
|
|
|
$links = $harvest->getLinks(Harvest::LINK_INTERNAL); |
|
185
|
|
|
|
|
186
|
9 |
|
//$url->ratio_text_code = $harvest->getRatioTxtCode(); // Slow ~30% |
|
187
|
|
|
$url->words_count = $harvest->getTextAnalysis()->getWordNumber(); |
|
188
|
|
|
$url->load_time = $harvest->getResponse()->getInfo('total_time'); |
|
189
|
6 |
|
$url->size = $harvest->getResponse()->getInfo('size_download'); |
|
190
|
|
|
|
|
191
|
6 |
|
/* |
|
192
|
|
|
* I remove it from the default crawler, you can extend this one and restablish this code |
|
193
|
|
|
* |
|
194
|
|
|
$breadcrumb = $harvest->getBreadCrumb(); |
|
195
|
6 |
|
if (is_array($breadcrumb)) { |
|
196
|
|
|
$url->breadcrumb_level = count($breadcrumb); |
|
197
|
|
|
$url->breadcrumb_first = isset($breadcrumb[1]) ? $breadcrumb[1]->getCleanName() : ''; |
|
198
|
9 |
|
$url->breadcrumb_text = $harvest->getBreadCrumb('//'); |
|
199
|
|
|
} |
|
200
|
9 |
|
* |
|
201
|
9 |
|
$url->kws = ','.implode(',', array_keys($harvest->getKws())).','; // Slow ~20% |
|
202
|
9 |
|
/**/ |
|
203
|
9 |
|
|
|
204
|
9 |
|
$url->title = $harvest->getUniqueTag('head title') ?? ''; |
|
205
|
|
|
$url->h1 = $harvest->getUniqueTag('h1') ?? ''; |
|
206
|
|
|
$url->h1 = $url->title == $url->h1 ? '=' : $url->h1; |
|
207
|
|
|
} |
|
208
|
9 |
|
|
|
209
|
|
|
if (isset($links)) { |
|
210
|
9 |
|
$this->updateInboundLinksCounter($url, $links, $harvest); |
|
211
|
9 |
|
$this->recorder->recordLinksIndex($this->config->getBase(), $url, $this->urls, $harvest->getLinks()); |
|
212
|
|
|
} |
|
213
|
9 |
|
|
|
214
|
|
|
return $harvest; |
|
215
|
9 |
|
} |
|
216
|
|
|
|
|
217
|
|
|
public function updateInboundLinksCounter(Url $url, array $links, Harvest $harvest) |
|
218
|
|
|
{ |
|
219
|
9 |
|
$everAdd = []; |
|
220
|
|
|
foreach ($links as $link) { |
|
221
|
9 |
|
$newUri = substr($link->getPageUrl(), strlen($this->config->getBase())); |
|
222
|
3 |
|
$this->urls[$newUri] = $this->urls[$newUri] ?? new Url($link->getPageUrl(), ($this->currentClick + 1)); |
|
223
|
|
|
if (!isset($everAdd[$newUri])) { |
|
224
|
3 |
|
$everAdd[$newUri] = 1; |
|
225
|
|
|
if (!$link->mayFollow() || !$harvest->mayFollow()) { |
|
226
|
|
|
++$this->urls[$newUri]->inboundlinks_nofollow; |
|
227
|
6 |
|
$this->recorder->recordInboundLink($url, $this->urls[$newUri], self::NOFOLLOW); |
|
228
|
|
|
} else { |
|
229
|
6 |
|
++$this->urls[$newUri]->inboundlinks; |
|
230
|
|
|
$this->recorder->recordInboundLink($url, $this->urls[$newUri], self::FOLLOW); |
|
231
|
6 |
|
} |
|
232
|
|
|
} |
|
233
|
|
|
} |
|
234
|
|
|
} |
|
235
|
|
|
} |
|
236
|
|
|
|
PiedWeb /
SeoPocketCrawler
